pipeline v1-2

Author: jinsoo

qdrant_db/.lock

@@ -2,7 +2,7 @@
 
 # Core dependencies
 transformers>=4.30.0
-#torch>=2.0.0
+torch>=2.0.0
 sentence-transformers>=2.2.0
 qdrant-client>=1.6.0
 #accelerate>=0.20.0
@@ -22,6 +22,7 @@ pandas>=2.0.0
 # Utilities
 python-dotenv>=1.0.0
 requests>=2.31.0
+httpx>=0.25.0
 psutil>=7.0.0
 
 # GPU Optimization (Phase 1 & 2)

qdrant_db/collection/restaurant_vectors/storage.sqlite

@@ -18,14 +18,10 @@
     prepare_qdrant_points,
     get_restaurant_reviews,
     query_similar_reviews,
-    get_reviews_with_images,
 )
-from .llm_utils import (
-    LLMUtils,
-    summarize_reviews,
-)
-from .strength_extraction import (
-    StrengthExtractionPipeline,
+from .llm_utils import LLMUtils
+from .comparison import (
+    ComparisonPipeline,
 )
 
 __all__ = [
@@ -36,9 +32,7 @@
     "prepare_qdrant_points",
     "get_restaurant_reviews",
     "query_similar_reviews",
-    "get_reviews_with_images",
     "LLMUtils",
-    "summarize_reviews",
-    "StrengthExtractionPipeline",
+    "ComparisonPipeline",
 ]
 

qdrant_db/collection/reviews_collection/storage.sqlite

@@ -12,10 +12,14 @@
 from ..vector_search import VectorSearch
 from ..llm_utils import LLMUtils
 from ..metrics_collector import MetricsCollector
+from ..cpu_monitor import get_or_create_benchmark_cpu_monitor
+from ..gpu_monitor import get_or_create_benchmark_gpu_monitor
 
 # 의존성 싱글톤 (매 요청 인스턴스 생성 방지 — Encoder/Qdrant는 @lru_cache, 나머지는 모듈 캐시)
 _vector_search_singleton: Optional[VectorSearch] = None
 _sentiment_analyzer_singleton: Optional[SentimentAnalyzer] = None
+_default_metrics_collector: Optional[MetricsCollector] = None
+_benchmark_metrics_collector: Optional[MetricsCollector] = None
 
 
 @lru_cache()
@@ -44,15 +48,56 @@ def get_qdrant_client() -> QdrantClient:
     return QdrantClient(path=qdrant_path)
 
 
-@lru_cache()
-def get_metrics_collector() -> MetricsCollector:
-    """메트릭 수집기 싱글톤"""
-    return MetricsCollector(
-        enable_logging=Config.METRICS_ENABLE_LOGGING,
-        enable_db=Config.METRICS_ENABLE_DB,
-        db_path=Config.METRICS_DB_PATH,
-        log_dir=Config.METRICS_LOG_DIR,
-    )
+def get_metrics_collector(
+    x_benchmark: Optional[str] = Header(None, alias="X-Benchmark"),
+    x_enable_cpu_monitor: Optional[str] = Header(None, alias="X-Enable-CPU-Monitor"),
+    x_enable_gpu_monitor: Optional[str] = Header(None, alias="X-Enable-GPU-Monitor"),
+) -> MetricsCollector:
+    """
+    메트릭 수집기 싱글톤.
+    - X-Benchmark: true 이면 요청 메트릭 수집 활성화 (logs + metrics.db).
+    - X-Enable-CPU-Monitor: true 이면 CPU 모니터링만 활성화 (logs/cpu_usage.log).
+    - X-Enable-GPU-Monitor: true 이면 서버 GPU 모니터링만 활성화 (logs/gpu_usage.log).
+    - 헤더는 독립적: 메트릭만 / CPU만 / GPU만 / 조합 가능.
+    """
+    global _default_metrics_collector, _benchmark_metrics_collector
+
+    _truth = ("true", "1", "yes")
+    want_cpu = x_enable_cpu_monitor and str(x_enable_cpu_monitor).strip().lower() in _truth
+    want_gpu = x_enable_gpu_monitor and str(x_enable_gpu_monitor).strip().lower() in _truth
+    want_metrics = x_benchmark and str(x_benchmark).strip().lower() in _truth
+
+    if want_cpu:
+        get_or_create_benchmark_cpu_monitor()
+    if want_gpu:
+        get_or_create_benchmark_gpu_monitor()
+
+    if want_metrics:
+        if _benchmark_metrics_collector is None:
+            _benchmark_metrics_collector = MetricsCollector(
+                enable_logging=True,
+                enable_db=True,
+                db_path=Config.METRICS_DB_PATH,
+                log_dir=Config.METRICS_LOG_DIR,
+            )
+        return _benchmark_metrics_collector
+
+    if _default_metrics_collector is None:
+        if not Config.METRICS_AND_LOGGING_ENABLE:
+            _default_metrics_collector = MetricsCollector(
+                enable_logging=False,
+                enable_db=False,
+                db_path=Config.METRICS_DB_PATH,
+                log_dir=Config.METRICS_LOG_DIR,
+            )
+        else:
+            _default_metrics_collector = MetricsCollector(
+                enable_logging=Config.METRICS_ENABLE_LOGGING,
+                enable_db=Config.METRICS_ENABLE_DB,
+                db_path=Config.METRICS_DB_PATH,
+                log_dir=Config.METRICS_LOG_DIR,
+            )
+    return _default_metrics_collector
 
 
 @lru_cache()

qdrant_db/meta.json

@@ -2,13 +2,18 @@
 FastAPI 메인 애플리케이션
 """
 
-from fastapi import FastAPI
+from fastapi import FastAPI, Request
 from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import JSONResponse
+from fastapi.exceptions import RequestValidationError
+from starlette.exceptions import HTTPException as StarletteHTTPException
 from contextlib import asynccontextmanager
 import logging
 import sys
+import uuid
 
-from .routers import sentiment, vector, llm, restaurant, test
+from .routers import sentiment, vector, llm, test
+from ..cpu_monitor import get_cpu_monitor
 
 # 로거 설정 (콘솔 출력)
 # basicConfig는 한 번만 실행되므로, root 로거에 직접 핸들러 추가
@@ -38,8 +43,17 @@ async def lifespan(app: FastAPI):
     """애플리케이션 생명주기 관리"""
     # 시작 시 초기화
     logger.info("FastAPI 애플리케이션 시작")
+    
+    # CPU 모니터 시작 (Config.CPU_MONITOR_ENABLE=true일 때만)
+    cpu_monitor = get_cpu_monitor()
+    if cpu_monitor:
+        cpu_monitor.start()
+    
     yield
+    
     # 종료 시 정리
+    if cpu_monitor:
+        await cpu_monitor.stop()
     logger.info("FastAPI 애플리케이션 종료")
 
 
@@ -50,6 +64,65 @@ async def lifespan(app: FastAPI):
     lifespan=lifespan,
 )
 
+# 요청 ID 미들웨어 (응답 헤더에도 포함)
+@app.middleware("http")
+async def add_request_id(request: Request, call_next):
+    request_id = request.headers.get("X-Request-Id") or request.headers.get("X-Request-ID") or str(uuid.uuid4())
+    request.state.request_id = request_id
+    response = await call_next(request)
+    response.headers["X-Request-Id"] = request_id
+    return response
+
+
+def _error_payload(*, code: int, message: str, details, request_id: str) -> dict:
+    return {"code": code, "message": message, "details": details, "request_id": request_id}
+
+
+@app.exception_handler(RequestValidationError)
+async def request_validation_exception_handler(request: Request, exc: RequestValidationError):
+    request_id = getattr(request.state, "request_id", str(uuid.uuid4()))
+    return JSONResponse(
+        status_code=422,
+        content=_error_payload(
+            code=422,
+            message="Validation error",
+            details=exc.errors(),
+            request_id=request_id,
+        ),
+    )
+
+
+@app.exception_handler(StarletteHTTPException)
+async def starlette_http_exception_handler(request: Request, exc: StarletteHTTPException):
+    request_id = getattr(request.state, "request_id", str(uuid.uuid4()))
+    # 404 등 라우팅 단계 HTTP 예외 포함
+    detail = getattr(exc, "detail", None)
+    message = str(detail) if detail is not None else "HTTP error"
+    return JSONResponse(
+        status_code=exc.status_code,
+        content=_error_payload(
+            code=exc.status_code,
+            message=message,
+            details=detail,
+            request_id=request_id,
+        ),
+    )
+
+
+@app.exception_handler(Exception)
+async def unhandled_exception_handler(request: Request, exc: Exception):
+    request_id = getattr(request.state, "request_id", str(uuid.uuid4()))
+    logger.error("Unhandled exception: %s", exc, exc_info=True)
+    return JSONResponse(
+        status_code=500,
+        content=_error_payload(
+            code=500,
+            message="Internal server error",
+            details=None,
+            request_id=request_id,
+        ),
+    )
+
 # CORS 설정
 app.add_middleware(
     CORSMiddleware,
@@ -63,7 +136,6 @@ async def lifespan(app: FastAPI):
 app.include_router(sentiment.router, prefix="/api/v1/sentiment", tags=["sentiment"])
 app.include_router(vector.router, prefix="/api/v1/vector", tags=["vector"])
 app.include_router(llm.router, prefix="/api/v1/llm", tags=["llm"])
-app.include_router(restaurant.router, prefix="/api/v1/restaurants", tags=["restaurants"])
 app.include_router(test.router, prefix="/api/v1/test", tags=["test"])